Re: [PATCH v2] spi: spi-geni-qcom: Speculative fix of "nobody cared" about interrupt

[Doug Anderson <dianders@xxxxxxxxxxxx>]

Hi,

On Mon, Mar 23, 2020 at 4:08 AM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>
> 561de45f72bd5f Girish Mahadevan 2018-10-03  327                 ret = get_spi_clk_cfg(xfer->speed_hz, mas, &idx, &div);
> 561de45f72bd5f Girish Mahadevan 2018-10-03  328                 if (ret) {
> 561de45f72bd5f Girish Mahadevan 2018-10-03  329                         dev_err(mas->dev, "Err setting clks:%d\n", ret);
> 561de45f72bd5f Girish Mahadevan 2018-10-03  330                         return;
>
> Needs to drop the lock before returning.

Oops, thanks for catching.  I will wait before spinning a v3 until
there is some clarity about whether folks want to do something more
like Stephen suggested or whether I should continue with my strategy.
At the moment I'm still in favor of keeping w/ my strategy and seeing
if I can reduce the amount of time with interrupts disabled in
setup_fifo_xfer(), maybe just grabbing the lock around the start of
the transfer to keep the state machine in sync with the kickoff...

-Doug